Method and device for discriminating cells by electromagnetic wave scattering spectrum

Abstract

Provided is a method for using a sample containing cells collected from a patient to discriminate, with high sensitivity and high accuracy, normal cells and abnormal cells contained in the sample. A sample containing cells collected from a patient is irradiated with electromagnetic waves, an electromagnetic wave scattering spectrum from the cells in the sample is measured, and normal cells and abnormal cells are discriminated on the basis of the characteristics of the electromagnetic wave scattering spectrum. The discrimination uses a trained model in which the electromagnetic wave scattering spectra of normal cells and abnormal cells are trained. The sample may be a living cell, or may be a sample that is immobilized by an alcohol or formalin and dyed. The electromagnetic wave scattering spectrum may use either or both of a scattering spectrum obtained through Rayleigh scattering and a scattering spectrum obtained through Mie scattering.

Description

Method and device for distinguishing cells using electromagnetic wave scattering spectrum

[0001] The present invention relates to a method and apparatus for easily and sensitively detecting abnormal cells (also called abnormal cells) that have degenerated from normal cells contained in a sample containing cells collected from a patient, with high accuracy, and in particular to a method and apparatus for distinguishing tumor cells, which are difficult to distinguish using an optical microscope.

[0002] One method for detecting malignant tumors is pathology, in which tissue sections from suspected patients are observed under an optical microscope. During pathology, a clinician diagnoses suspected malignant tumors. If a suspected malignant tumor is present, a pathologist receives pathology specimens, such as histological and cytological specimens, from the clinician to diagnose the tumor type and condition, and then develops a treatment plan based on that diagnosis. Thus, pathology plays an extremely important role in the medical field. However, compared to other developed countries, Japan has a low number of pathologists, and a small number of pathologists are responsible for diagnosing a large number of specimens. Furthermore, pathology diagnosis using an optical microscope often has difficulty distinguishing between tumor cells and normal cells, making diagnosis dependent on the pathologist's experience. In response to these issues, artificial intelligence (AI) analysis based on machine learning of optical microscope images of pathology specimens has recently attracted attention as a solution to these problems.

[0003] Even experienced pathologists may be unable to distinguish tumor cells from normal cells by observing specimens using an optical microscope alone. The differences between tumor cells and normal cells are not only apparent in the morphology of micrometer-sized cells and intracellular organelles observed under an optical microscope, but also in submicrometer-sized (nano-sized) cytoskeleton and microstructure within intracellular organelles, as well as protein denaturation. If this nano-sized information could be detected and analyzed using light scattering spectra in addition to optical microscope images, it is believed that the diagnostic accuracy of pathological specimens could be improved. Furthermore, if high-density information could be obtained faster than with optical microscope observation, it is expected that high-speed and high-precision AI analysis could be realized.

[0004] Analysis presupposes the preparation of appropriate samples and the use of accurate testing techniques. In cytology, the collected sample is typically smeared onto a glass slide and immediately dehydrated and fixed in ethanol. If the sample is dry, it is rehydrated and then fixed in 95% ethanol. After fixation, staining is performed, such as Papanicolaou staining. In pathology, the collected sample is typically fixed in formalin and paraffin-embedded into a block. Sections are then prepared and stained with hematoxylin and eosin (HE). If a tumor is suspected, immunostaining techniques corresponding to various tumor markers are used, allowing a pathologist to accurately identify the tumor. The stained sample is then observed to identify tumor cells based on characteristics such as enlarged nuclei, multinucleation, and abnormalities in the nuclear microstructure. However, there are many types of tumor cells that are extremely difficult to distinguish due to these differences. For example, in cytological examinations of malignant mesothelioma, a rare cancer, even specialists often have difficulty distinguishing mesothelioma cells from reactive mesothelial cells and adenocarcinoma cells, and more appropriate methods for this distinction are needed.

[0005] Because the pathology specimen preparation process (a series of procedures including fixation and staining) is extremely laborious and time-consuming, it is desirable to identify tumors during surgery or immediately after collection, before fixation and staining. Patent Document 1, for example, discloses a method for detecting abnormal cells (tumor cells) by directly irradiating live cells with laser light without staining them and spectroscopically analyzing the stimulated emission light emitted from the live cells. Specifically, the method aims to detect abnormal cells by irradiating cells with a visible laser and spectroscopically analyzing (spectrometry) the stimulated emission light from the biological sample. While the present invention utilizes light scattering spectra, in the examples of the present invention, the scattered light component of the laser light irradiated on the live cells is filtered out, and light scattering, which is the subject of the present invention, is not considered as detection information. The stimulated emission spectrum has low accuracy in identifying intracellular conditions, and some types of tumor cells may not be detected.

[0006] Alternatively, there are methods for staining live cells during surgery or immediately after collection without fixation, such as fluorescent staining using 5-aminolevulinic acid (5-ALA) as described in Non-Patent Document 1. 5-ALA is specifically taken up by tumor cells, metabolized (converted) into the fluorescent molecule protoporphyrin IX within the tumor cells, and then accumulated. This method of detecting the fluorescence of protoporphyrin IX to identify tumor cells is commonly used in the treatment of skin cancer in Europe and the United States, and in Japan, it is used as an intraoperative fluorescent diagnostic agent for brain tumors and bladder cancer to remove the tumor without leaving any residual tissue. However, this method, which identifies tumor cells solely based on the presence or absence of fluorescence from protoporphyrin IX, is difficult to identify tumor cells with a higher accuracy than observing pathological specimens stained with a wide variety of dyes, and it is unable to distinguish between tumor cell types. Furthermore, a staining method utilizing the metabolism of living tumor cells by 5-ALA cannot be applied to staining fixed specimens, and therefore cannot be used as a method for easily detecting normal cells and abnormal cells (tumor cells) contained in a pathological specimen containing cells collected from a patient with high sensitivity and accuracy, which is the objective of the present invention.

[0007] Various methods have also been investigated to distinguish between normal and abnormal cells (tumor cells) by directly irradiating living cells with electromagnetic waves such as light without staining them and using the scattered electromagnetic wave spectrum, but it has not been possible to distinguish between tumor cells with a higher accuracy than by observing pathological specimens.

[0008] For example, examples of detecting tumor cells by irradiating visible light and analyzing scattered light reflected from the surface of biological tissue and observed as backscattered light have been reported in Non-Patent Documents 2 and 3. Non-Patent Document 2 shows that enlarged nuclei, a characteristic of tumor cells, can be detected by analyzing the angular dependence of the intensity of Mie scattered light scattered back from living cells, which is derived from the size of the nuclei of the living cells. Similarly, Non-Patent Document 3 shows that the presence of enlarged nuclei in tumor cells can be detected by analyzing the wavelength dependence of the intensity of Mie scattered light scattered back from living cells. As in these examples, the changes in the scattering direction and spectral shape of Mie scattered light and Rayleigh scattered light due to the size of the nuclei of cells with simple, nearly spherical shapes can be theoretically predicted based on physics. However, it is impossible to unambiguously predict the changes in the scattering direction and spectral shape of Mie scattered light and Rayleigh scattered light due to the more complex shapes within the cells, making it impossible to distinguish between tumor cell types.

[0009] Patent Document 2 discloses a method for identifying cell types by using a flow cytometer to measure a fluid containing a mixture of various cell types, and analyzing the time-series signals of optical images or spectra obtained while varying the relative position of the illumination pattern and the object being observed in a microchannel using AI based on machine learning, which does not rely on a definitive physical prediction. This method allows for the distinction between normal and abnormal cells in a large number of cells to be analyzed using AI based on machine learning. However, its application is limited to cases in which time-series signals are obtained from cells flowing at a constant speed, as in a flow cytometer, and it is difficult to use it with immobilized cells adhered to a substrate, such as pathology specimens. However, the present invention is characterized by not requiring a time-series signal. Therefore, the invention described in Patent Document 2 cannot be used as a method for easily detecting normal and tumor cells contained in a stationary sample containing cells collected from a patient, as is the purpose of the present invention, with high sensitivity and accuracy.

[0010] As mentioned above, various methods for measuring living cells have been used to distinguish between normal and abnormal cells, primarily for the purpose of identifying the extent of tumor tissue during surgery. However, these methods do not utilize more detailed methods for distinguishing between tumor cells and tissues containing tumor cells, as they do not utilize tumor markers that are expressed in various types of tumors.

[0011] The present invention aims to provide a method for distinguishing between normal and abnormal living cells, and also to provide a method for distinguishing between normal and abnormal cells in fixed and stained pathological specimens prepared according to existing pathological diagnostic processes, by measuring the light scattering spectrum of the pathological specimen based on nanometer to subnanometer morphological characteristics such as the microscopic shape of the cells, the shape of the nucleus, and the intranuclear microstructure, and analyzing the light scattering spectrum using machine learning. Specifically, the present invention aims to provide a method for easily detecting normal and abnormal cells contained in a sample containing cells collected from a patient with high sensitivity and accuracy. The sample may be a pathological specimen collected from a patient and subjected to fixation and staining processes in accordance with existing pathological diagnostic procedures, or a specimen prior to the fixation and staining processes. The present invention aims to provide a means for easily performing diagnosis with high sensitivity and accuracy, without relying on the experience and skill of a pathologist, rather than relying on the experience and skill of a pathologist. In particular, the present invention aims to provide a method for distinguishing between tumor cells, which are difficult to distinguish using an optical microscope.

[0012] JP-A-58-118948 Patent No. 6959614

[0013] S. Ogura et al., “Photodynamic therapy targeting dormant cancer cells using 5-aminolevulinic acid,” Journal of the Japan Society for Laser Surgery and Medicine, Vol. 43, No. 4, pp. 238-248 (2023). A. Wax et al., “Cellular tissue and substructure measured using angle-resolved low-coherence interferometry,” Biophysical Journal, 82(4), pp. 2256-2264 (2002). V. Backman et al., “Polarized light scattering spectroscopy for quantitative measurement of epithelial cellular structures in situ,” IEEE Journal of Selected Topics in Quantum Electronics, 5(4), pp. 1019-1026 (1999).

[0014] The present invention aims to provide a method and apparatus for detecting, with high sensitivity and accuracy, normal and abnormal cells contained in a sample containing cells collected from a patient, particularly tumor cells that are difficult to distinguish using an optical microscope.

[0015] To address the above-mentioned challenges, the cell discrimination method of the present invention irradiates electromagnetic waves onto a cell-containing sample (cell sample) collected from a patient, obtains an electromagnetic wave scattering spectrum from the sample, and discriminates between normal and abnormal cells based on the information obtained from the spectrum. Light scattering is affected by the reflection and refraction of structures that are the same size as or smaller than the wavelength of light (100 nm to 1 μm). Light scattering by structures on the same size as the wavelength (micrometer scale) is called Mie scattering, while light scattering by structures smaller than the wavelength (submicrometer scale) is called Rayleigh scattering. The present invention is characterized by utilizing the information (spectral shape) contained in the spectral spectra of Rayleigh scattering and Mie scattering to identify abnormal cells. While optical images also contain information from Rayleigh scattering and Mie scattering, optical images do not contain information beyond the diffraction limit. By analyzing the light scattering spectrum, a feature of the present invention, information on submicron-scale intracellular structures that cannot be observed with optical microscopes beyond the diffraction limit can be obtained. What is novel about this method is that information is extracted from the light scattering spectrum using a mathematical method rather than relying on physical or chemical interpretations. When a cell sample is irradiated with light, the scattering direction and spectral shape of the light change depending on the micrometer-scale shapes of the cells and cell nuclei, as well as the submicrometer-scale microstructure of the cells, the shape of the nuclei, and the microstructure within the nuclei. Furthermore, when cells are stained, as in pathology specimens, and processed with dyes to highlight the structural features of abnormal cells and the types of proteins they contain, light scattering is strongly influenced by the dyes. The light scattering spectrum from a pathology specimen is thought to comprehensively reflect all of this information. It is therefore impossible to unambiguously link simple spectral features, such as undulations, observed in the light scattering spectrum with cellular and intracellular structures, or to theoretically analyze the spectrum based on physics. To address this issue, the present invention proposes a method for analyzing light scattering spectra using machine learning to predict pathology diagnosis results. Machine learning involves numerically extracting features of the light scattering spectrum, statistically analyzing these features, and training a predictive model (also known as a learning model) without relying on physical logic.AI analysis is a method of inputting a different light scattering spectrum into this prediction model to predict whether a cell is normal or abnormal. The numerically determined feature values ​​do not necessarily match the spectral characteristics perceived by humans visually, and may be determined by subtle spectral differences that cannot be discerned visually, resulting in an interpretation that is fundamentally different from traditional theoretical analysis based on physics. In other words, the present invention solves the problem of interpreting light scattering spectrum data, in which the causal relationship between cells and their microstructures, which reflects the difference between normal and abnormal cells, is too complex to be analyzed theoretically based on physics, by using machine learning, a numerical analysis not based on physical logic.

[0016] The present invention has the effect of providing a method for detecting normal cells and abnormal cells with high sensitivity and accuracy. It can facilitate the discrimination of tumor cells, which have been difficult to discriminate between in the conventional diagnosis by a pathologist using optical microscope observation, and can support diagnosis that does not depend on the experience or skill of the pathologist, and can also support the improvement of the accuracy of that diagnosis.

[0017] Diagram showing an overview of the measurement system (a) Overall configuration of the measurement system (b) Enlarged view of the illumination section for dark-field image observation (c) Enlarged view of the illumination section for bright-field image observation Configuration of MDCK cell sample Comparison of bright-field and dark-field images of MDCK cells without (a) and with (b) Cytochalasin D added (a) Normalized light scattering spectra (SES) immediately after the start of measurement and 60 minutes after measurement from MDCK cells immersed in culture medium ｎ７１０ (b) Comparison of normalized light scattering spectra (SES) from cells every 10 minutes after adding Cytochalasin D to the cells for 60 minutes. ｎ７１０) Changes in the light scattering spectra (SES) of cells in specimens A to F from normal patients. Bright-field and dark-field images of cells in specimens A to F from normal patients. (a) Cells with a peak observed in the SES around 700 nm. (b) Cells with a flat SES. Light scattering spectra (SES) of cells in specimens G, H, and I from patients suspected of mesothelioma. Bright-field and dark-field images of cells in specimens G, H, and I from patients suspected of mesothelioma. (a) The first component obtained by principal component analysis of the light scattering spectra. Loadings of the first and second principal components, and (b) scatter plot of the percentage ratio (score) of the first and second principal components contained in the light scattering spectrum. Conceptual diagram of two-class classification using k-nearest neighbor (a), decision tree (b), and random forest (c). Diagram explaining how to divide the training data and test data for specimens A to I to verify the classification accuracy (correct rate) of light scattering spectra using a machine learning model. Block diagram of the cell discrimination device. Normalized light scattering spectra (SES) for cells contained in specimens A to I. ｎ７１０ ) Contribution ratios of the first principal component (PC1) to the tenth principal component and SES of the loading result cells ｎ７１０ Scatter plot of the scores of the nth and mth principal components. Illustrative diagram of the malignancy of tumor cells. Illustrative diagram of support for pathological diagnosis. Normalized light scattering spectra (SES) for two specific types of living cells. ｎ７１０ ) SES of two specific types of cells ｎ７１０ Scatter plot of scores for the nth and mth principal components. Illustrative diagram of the scattered light spectrum measurement method.

[0018] The present invention involves irradiating a sample containing cells collected from a patient with electromagnetic waves, obtaining a light scattering spectrum of the scattered light generated, and then numerically extracting the features contained in the light scattering spectrum to distinguish between normal and abnormal cells based on the differences in the characteristics of their light scattering spectra, creating a learning model based on the features, and discriminating between unknown cells.

[0019] Cells contain various organelles, including the nucleus, within the cytosol, which is separated from the outside by the cell membrane. It is known that characteristic changes occur in the size and fine structure of various organelles between normal and abnormal cells. The present invention is intended to capture the characteristics of abnormal cells, including various tumor cells that occupy an important position in pathological diagnosis, using the feature quantities of light scattering spectra. However, the present invention is not limited to these, and provides a method for detecting and analyzing various intracellular and extracellular fine structures by utilizing light scattering spectrum measurement.

[0020] Degeneration of cells and subcellular structures can occur for a variety of reasons, including aging, chemical and physical damage, and stimulation by various physiologically active substances, infection, and cancer. Early detection of cellular abnormalities is crucial through the morphological changes of the nucleus and other organelles contained within. Cell degeneration, particularly cell shape and nuclear degeneration, plays an important role in various pathological diagnoses. For example, tumor cells may exhibit characteristics such as cell clumping, enlarged and multinucleated nuclei, enlarged nucleoli, and narrowed cytosol. Furthermore, tumor cells may exhibit abnormal polymerization of cytoskeletal proteins such as actin and keratin.

[0021] Furthermore, many malignant tumors become more mobile and undergo epithelial cell degeneration (epithelial-mesenchymal transition), leading to infiltration of the interstitium beyond the basement membrane and metastasis to other organs. During this process, reorganization of actin filaments, which maintain the cytoskeleton, occurs, and filopodia are formed on the cell membrane, increasing cell mobility. Therefore, cell degeneration due to tumors and other conditions can be detected by changes in cell shape, the shape and size of the cell nucleus, intracellular organelles, and the cytoskeleton, which are changes in shape on the micrometer to submicrometer scale. The present invention uses a cell discrimination method in which electromagnetic waves are irradiated onto a sample containing cells collected from a patient, electromagnetic scattering spectra from the cells in the sample are acquired, and a learning model trained to identify the characteristics of the electromagnetic scattering spectra of normal and abnormal cells is used to discriminate between normal and abnormal cells. The sample may be live cells, or may be a sample fixed and stained with alcohol or formalin.

[0022] There has been extensive research into capturing cellular characteristics using light scattering. The light scattering phenomenon is characterized by the particle size parameter (α; α = πd / λ), which is the ratio of the diameter (d) of the scatterer to the wavelength (λ) of the electromagnetic wave being observed, and the relative refractive index of the scatterer. When α << 1, Rayleigh scattering predominates, and the scattered light intensity increases as the wavelength shortens. When α is close to 1, Mie scattering predominates, and the scattered light intensity assumes a constant value independent of α. Furthermore, when α is 10 or greater, diffraction occurs due to the interference of scattered light, resulting in complex behavior with respect to the scattering angle and α. In the present invention, the scattering spectrum can be either a Rayleigh scattering spectrum or a Mie scattering spectrum, or both. More preferably, the scattering spectrum can be either a Rayleigh scattering forward scattering spectrum or a Mie scattering forward scattering spectrum, or both. Furthermore, the difference between the electromagnetic scattering spectra of normal and abnormal cells is due to the increase or decrease in scattering caused by changes in cellular structure that are smaller or equal to the wavelength of visible light.

[0023] The wavelength of the electromagnetic waves used for observation in the present invention is preferably 100 nm to 1 μm, with visible light in the wavelength range of 400 to 750 nm being particularly preferred. Light in the range from ultraviolet to infrared can also be used. Furthermore, when using the present invention, it is also possible to use electromagnetic waves in the wavelength range from X-rays to ultraviolet to obtain electromagnetic scattering spectra from even smaller regions of cells in a sample, and to use a learning model that has learned the characteristics of the electromagnetic scattering spectra of normal and abnormal cells to discriminate between normal and abnormal cells.

[0024] The present invention is characterized by detecting the structural characteristics of scatterers with high sensitivity and precision by observing scattered light from individual cells. In particular, because scatterers include structures such as nuclei and cytoskeleton that constitute cells, the range of α includes 0.1 < α < 150. Because the size of cell nuclei ranges from several micrometers to approximately 10 micrometers, when the wavelength of the electromagnetic waves being observed is approximately 500 nm, the value of α is approximately 20 to 120. In this case, the scattered light intensity is highly dependent on the scattering angle, and most of the scattering is observed as forward scattering in the same direction as the incident light. When using the method of the present invention, the scattered light from cells is observed by detecting scattered light scattered in any direction relative to the incident angle of the electromagnetic waves irradiated onto the cells, thereby obtaining an electromagnetic scattering spectrum from the cells in the sample. A learning model that has been trained to learn the characteristics of the electromagnetic scattering spectra of normal and abnormal cells can be used to perform a cell discrimination method for distinguishing between normal and abnormal cells.

[0025] When the cross-sectional area of ​​the scatterer is larger than the wavelength of the observed electromagnetic wave, the scattering efficiency varies in a complex manner with α. It is known that, when the scattering efficiency has the relative refractive index of cellular components in a medium such as water, it significantly depends on the size of the scatterer and the wavelength of the electromagnetic wave. Non-Patent Document 4 (M. Xu et al., "Unified Mie and fractal scattering by cells and experimental study on application in optical characterization of cellular and subcellular structures," Journal of Biomedical Optics, 13(2), 024015-024015 (2008)) presents the results of comparing the light scattering spectra measured by irradiating visible light onto cells and varying the scattering angle between 1.1 and 165° with calculated values. Furthermore, the scattered light from cells is expressed as a superposition of scattered light components from the nucleus, homogeneous cytoplasm, and minute organelles other than the nucleus. The light scattering spectra measured at various scattering angles are compared with calculated values. According to this, for forward scattering at a scattering angle of 1.1°, the scattered light intensity increases with wavelength, while for backward scattering, the phase is reversed, and the scattered light intensity tends to decrease with wavelength. It can be seen that the contributions of each component of scattered light include background scattering from the cell body cytoplasm, fractal "fluctuation" components from organelles, and scattering components from the nucleus, which contribute significantly. As described above, it has been demonstrated that accurate measurement of the light scattering spectrum from cells is possible when irradiating cells with visible light at scattering angles ranging from 1.1 to 165°. The method of the present invention can also be used as a cell discrimination method, obtaining electromagnetic scattering spectra from cells regardless of the observation conditions for forward scattering or backward scattering, and using a learning model that has learned the characteristics of the electromagnetic scattering spectra of normal and abnormal cells to discriminate between normal and abnormal cells.

[0026] In Non-Patent Document 4, as the size of the nucleus increases, the dependence of scattered light intensity on wavelength becomes more pronounced in forward scattering, increasing significantly with wavelength, while the opposite trend is observed in backward scattering. Non-Patent Document 4 also examined the scattering behavior when cells were denatured by the action of acetic acid. It was found that the acetic acid denatured the organelles, causing whitening, but the scattering intensity increased significantly, and the contribution of the "fluctuation" component increased as the wavelength became shorter. As such, the present invention makes it possible to capture and analyze scattered light from both forward and backward scattering angles, thereby making it possible to accurately identify the differences between normal and abnormal cells that were previously unclear using various conventional methods.

[0027] In the present invention, the light scattering spectrum obtained by measurement is expressed as two-dimensional data representing the relationship between scattered light intensity and wavelength. Since the present invention collects and analyzes a large number of light scattering spectra obtained for individual cells, principal component analysis (PCA) was used to extract features from the two-dimensional data. Principal components were extracted by diagonalizing the variance-covariance matrix of the scattered light intensity at each wavelength. The first principal component is the eigenvector with the largest eigenvalue, which maximizes the variance of the scattered light intensity. Using the first principal component and the second principal component, which is the eigenvector with the second largest eigenvalue, the characteristics of the scattered light from the cells were analyzed, and a prediction model was created to classify the differences in the features of normal and abnormal cells. Specifically, the k-Nearest Neighbors algorithm (kNN), decision tree analysis, and random forest analysis (RF) were used as prediction models. For accuracy evaluation, the trained prediction model was used to predict whether a cell was normal or abnormal based on the features of the light scattering spectra of normal and abnormal cells that were not used in training, and the validity and accuracy of the analysis method were evaluated based on the accuracy rate.

[0028] The present invention provides an AI analysis method for measuring the light scattering spectrum of individual cells in a sample containing cells collected from a patient, analyzing the features of the obtained light scattering spectrum using the above-mentioned method, and using a learning model constructed from the previously acquired light scattering spectrum to determine whether the cells contained in the sample are normal or abnormal. Furthermore, when performing such AI analysis, it is preferable to use a cell discrimination device 20, as shown in FIG. 12 , which includes a scattering spectrum acquisition unit 21 that irradiates electromagnetic waves onto a sample containing cells collected from a patient and acquires a scattering spectrum 32 in an electromagnetic wave scattering spectrum measurement device 31, a learning model 22 that has previously learned the scattering spectra of normal cells and abnormal cells, and a discrimination unit 23 that uses the learning model 22 to discriminate between normal and abnormal cells. The discrimination results of the discrimination unit 23 of the cell discrimination device 20 are output to a display device 33, such as a display that can numerically classify and display normal cells and abnormal cells (tumor cells). The cell discrimination device 20 is a computer, and the learning model 22 and programs for the scattering spectrum acquisition unit 21 and discrimination unit 23 are stored in memory, and the programs are executed by a processor. An example of an embodiment of the present invention will be described in detail below with reference to the drawings. Note that the scope of the present invention is not limited to the following examples and illustrated examples, and many modifications and variations are possible.

[0029] In this example, experiments using cultured animal cells demonstrate that changes in submicrometer intracellular structures, difficult to observe using an optical microscope, are causally related to light scattering spectra, and that light scattering spectra contain information about the nuclear microstructure. Light scattering spectra for single cells were measured using a dark-field inverted microscope coupled with a spectrometer over a wavelength range of 420 to 720 nm. In this example, the polymerization inhibitor Cytochalasin D was added to cultured animal cells to fragment the fibrous actin filaments that make up the cytoskeleton. Experiments were conducted to determine the changes in the light scattering spectra of single cells. Because the light scattering spectrum from a cell contains the shape of the white light spectrum from the light source, the scattering efficiency spectrum (SES), calculated by dividing the actual scattered light measurement by the scattered light intensity from the light source, was used as the standard for the light scattering spectrum. SES indicates the wavelength dependence of scattered light intensity when light of equal intensity is incident at all wavelengths. The absolute value of SES varies significantly from cell to cell, depending on factors such as the cell's condition, the light irradiation conditions, and the adjustment of the objective lens used to collect scattered light. To correct for the variations due to this measurement, the entire spectrum was normalized by the SES value at 710 nm to obtain the normalized light scattering spectrum (SES ｎ７１０ ) was used for machine learning analysis.

[0030] A canine kidney tubular epithelial cell line (MDCK; RCB0995, RIKEN BioResource Center) was used. The cells were cultured on cell culture dishes in Dulbecco's modified Eagle's medium (phenol red-free) supplemented with fetal bovine serum (FBS) and antibiotics. Just before the cells reached confluence (80% coverage), the medium was replaced and the cell culture dish was placed on the sample stage. The center of the cell culture dish was observed using a darkfield inverted microscope (Olympus IX81 equipped with a darkfield condenser lens). A ring-shaped spacer 11 (30 μm thick) was placed in the center of the cell culture dish 10. MDCK cells 14 were seeded and cultured in a 1.70 cm diameter area within the ring, filled with medium 12. A cover glass 13 was placed on top to prevent the medium from drying out during observation.

[0031] Figure 1 shows an overview of the measurement system. Figure 2 shows the configuration of the cell culture dish used for the measurements. As shown in Figure 1(b), a halogen lamp was used as the illumination light source. The illumination light was irradiated onto the cells on the cell culture dish through a dark-field condenser lens, and scattered light from the cells was collected by an objective lens. Because the dark-field condenser lens illuminates the cells with a numerical aperture larger than that of the objective lens, the light collected by the objective lens does not contain any components of light that has passed through the cells (transmitted light). On the other hand, a normal optical microscope image (bright-field image) is obtained by collecting light that has passed through the cells with an objective lens, as shown in Figure 1(c). The scattered light collected by the objective lens is split into two paths, one of which is focused on a CMOS camera to obtain a dark-field image. The other path is introduced into a spectrometer via an optical fiber. The scattered light intensity spectrum measured by the spectrometer is divided by the spectrum of the illumination light to obtain the SES. In the scattering image formed on the optical fiber side, the core diameter of the optical fiber corresponds to an area of ​​approximately 30 μm in diameter on a cell culture dish, so SES of approximately one cell can be selectively measured. Here, the spectrometer detects Raman scattering in addition to Rayleigh scattering and Mie scattering. However, the intensity of Raman scattering is overwhelmingly lower than that of Mie scattering and Rayleigh scattering, and is below the detection sensitivity of the spectrometer, so that it is negligible or not detected in this measurement. That is, as shown in Figure 20 (1), in the present invention, the sample (scatterer) is irradiated with light containing all wavelengths in the visible range, such as white light, as excitation light 41. The scattered light, Mie scattering and Rayleigh scattering 43, are overwhelmingly stronger in intensity than Raman scattering and fluorescence 44. Therefore, Raman scattering and fluorescence 44 are ignored as scattering spectral components in the spectrometer 45. On the other hand, as shown in Figure 20 (2), in an optical system for selectively detecting Raman scattered light or fluorescence 44, light having a high intensity at a specific wavelength is used as excitation light 46, and unless a powerful optical filter (excitation light cut filter) 47 that cuts out Mie scattered light and Rayleigh scattered light is used, the main components of the detected light scattering will be Mie scattering and Rayleigh scattering. Note that Mie scattered light and Rayleigh scattered light have the same wavelength as the irradiating light (excitation light) that irradiates the cells, but Raman scattered light and fluorescence have wavelengths different from that of the excitation light.

[0032] Figure 3 shows bright-field and dark-field images of MDCK cells observed with a CMOS camera. Figure 3(a) shows an image of an MDCK cell without the addition of cytochalasin D, while Figure 3(b) shows an image of an MDCK cell with the addition of cytochalasin D and fragmented cytoskeleton. Regarding the bright-field and dark-field images in Figures 3(a) and (b), no significant differences were visible between the images immediately after (0 minutes) and 60 minutes after the addition of cytochalasin D. The actual fragmentation of actin filaments caused by the addition of cytochalasin D under the above conditions was confirmed by observing fluorescently stained (SPY555-actin, SPIROCHROME, SC202) samples using a confocal microscope. Actin filaments, which make up the cytoskeleton, do not absorb light in the visible light range and have a diameter of less than 10 nm, making them difficult to observe with an optical microscope. Therefore, it is reasonable that actin filament fragmentation cannot be observed in the bright-field images on the left side of Figures 3(a) and (b). While actin filament fragmentation may be observed in the dark-field images on the right side of Figures 3(a) and (b), no clear difference can be discerned visually. However, by applying machine learning analysis (described below), it may be possible to quantify the structural changes and classify cells with fragmented actin filaments.

[0033] Figure 4(a) shows a comparison of SES from MDCK cells in a medium without Cytochalasin D at 0 and 60 minutes after microscopy. ｎ７１０ Even without adding cytochalasin D, the structure of the cells changes when left for a long time, and the SES ｎ７１０ However, under these experimental conditions, this experiment showed that there was no change in the cell structure that could be seen on the light scattering spectrum for at least 60 minutes.

[0034] Next, MDCK cells were treated with cytochalasin D at a concentration of 200 μM to inhibit actin filament polymerization. SES was measured every 10 minutes from immediately after the addition until 60 minutes later. ｎ７１０ The change in the scattering intensity is shown in Figure 4(b). The results in Figure 4(b) indicate that changes in the light scattering spectrum were observed immediately after the addition of Cytochalasin D. These results indicate a tendency for the light scattering intensity to increase over time at wavelengths shorter than 500 nm. It is believed that the inhibition of actin filament polymerization contributes to Rayleigh scattering by actin filament fragments of various lengths, and this contribution can explain the increase in scattering intensity at shorter wavelengths seen in Figure 4(b). This example demonstrates that the visible light scattering spectrum is causally related to intracellular structural changes and can even measure differences in intracellular structure that cannot be clearly distinguished using bright-field or dark-field images. As shown in Figure 4, the visible light scattering spectrum of unstained, transparent cells exhibits a flat spectrum with no particularly clear features. However, this spectrum includes not only light scattering from the actin filaments, which are the cytoskeleton, but also light scattering from the fine structure of the cell, the nucleus, and fine structures within the nucleus. It is believed that the overall reflection of all these light scattering effects results in a spectrum that appears flat overall, but the slight irregularities that appear like noise on the flat spectrum may contain information that is causally related to these fine structures. However, it is nearly impossible to predict these irregularities unambiguously from a physical perspective. Therefore, as a solution to this problem, the present invention attempts to analyze information attributable to intracellular fine structures from light scattering spectra by analyzing them using AI based on machine learning, which does not rely on unambiguous physical predictions.

[0035] In this example, we use pathological specimens obtained from actual patients, in which normal and tumor cells were fixed and stained according to existing pathological diagnostic procedures. Results demonstrate that machine learning can be used to distinguish light scattering spectra, which are extremely difficult to interpret from a physical perspective. Pleural fluid was collected from the upper rib margin of each normal specimen (A-F) and specimens containing suspected pleural mesothelioma cells (G, H, I) using a syringe. Cells contained in the pleural fluid were smeared onto a glass slide, spray-fixed, and then fixed in 95% alcohol fixative. Following the standard Papanicolaou staining method, cell nuclei were stained with hematoxylin, and the cytoplasm was stained with OG-6 staining solution, followed by EA-50 staining. The specimen was then cleared in a xylol bath and finally mounted. The cells in the specimen were identified by a specialized pathologist, and the light scattering spectra were measured using specimens in which normal pleural mesothelial cells and other tumor cells were distinguished and individually marked.

[0036] For normal cells, 23 to 25 pre-marked normal cells contained in specimens A to F were measured, and for tumor cells, 13 to 27 pre-marked tumor cells contained in specimens G to I were measured. Here, tumor cells include not only mesothelioma cells but also adenocarcinoma cells. Furthermore, the possibility cannot be ruled out that the cells classified as normal cells and tumor cells may also contain reactive mesothelial cells.

[0037] One of the characteristics of mesothelioma cells is that they are more likely to have two nuclei per cell than other adenocarcinoma cells or reactive mesothelial cells. Furthermore, mesothelioma and adenocarcinoma cells are characterized by their tendency to form cell clusters, making them distinguishable from isolated reactive mesothelial cells. However, while mesothelioma cells are intermediate between adenocarcinoma and reactive mesothelial cells when simply comparing nuclear size, there is a distribution of nuclear size, making it difficult to distinguish between them without observing the distribution of nuclear size across multiple cells rather than just observing individual cell nuclei. While mesothelioma and adenocarcinoma nuclei are centrally located within the cell, adenocarcinoma cells are often ubiquitous at the cell edges. However, nuclear location alone does not allow for differentiation from reactive mesothelioma cells.

[0038] In this example, the purpose is to verify that the light scattering spectrum from tumor cells differs from that of normal cells, and not to distinguish mesothelioma cells from adenocarcinoma cells or reactive mesothelial cells. In other words, whether cells in a specimen are mesothelioma cells or normal cells is determined from both pathological and clinical perspectives, and the purpose of the present invention is to detect abnormal cells that differ from normal cells.

[0039] Figure 5 shows the light scattering efficiency spectra (SES) measured for cells contained in normal specimens A to F. The SES characteristics for the cells contained in each specimen differed, with the SES from cells in specimens A and F showing roughly flat spectra, while cells from specimens B, C, and D were characterized by the inclusion of cells that showed a peak in scattering efficiency around 700 nm. In the case of specimen E, a peak around 700 nm was observed in the SES for only some of the cells measured, but the rest showed roughly flat spectra.

[0040] Figure 6 shows the bright-field (top) and dark-field (bottom) images of cells in which a scattering peak was observed around 700 nm when SES was measured for the cells in specimens A to F. (a) shows the bright-field (top) and dark-field (bottom) images of cells in which the SES was flat. For the cells in specimen B, it was confirmed that the nuclei in which the scattering peak was observed were large, but for the cells in specimen C, no clear correlation was observed between the nuclear size and the scattering peak. This indicates that the appearance of the scattering peak may be influenced by factors other than the enlargement of the nucleus, such as differences in the relative refractive index due to the ease of staining.

[0041] Next, the results of measuring SES for cells contained in specimens G to I suspected of being mesothelioma are shown in Figure 7. Optical microscope images of cells in each specimen (bright-field image at the top, dark-field image at the bottom) are shown in Figure 8.

[0042] In Figure 6, all cells in specimen G exhibited flat SES, whereas a small number of cells in specimen H exhibited SES peaks around 700 nm. In specimen I, the majority of cells exhibited SES peaks around 700 nm. Furthermore, the optical microscope images of cells shown in Figure 8 revealed a distribution in the size of the nuclei contained within each cell, with some nuclei being centrally located and others being ubiquitous. These results demonstrate that SES clearly distinguishes between cells exhibiting a peak around 700 nm and flat cells, but these results alone do not necessarily indicate that normal cells can be distinguished from tumor cells. The SES from this cytological specimen likely reflects the causal relationships between the reflection, absorption, and refraction of the dyes bound to these cells by Papanicolaou staining, in addition to the fine structure of the cells, cell nuclei, cytoskeleton, and intracellular organelles. It is virtually impossible to analyze all of these SES considerations using a physics-based theory. Therefore, in this study, we used machine learning, a numerical analysis method not based on physical logic, as a method for discrimination.

[0043] Normalized light scattering spectra (SES) for the cells contained in specimens A to I ｎ７１０ ) was subjected to principal component analysis, the loadings of the first and second principal components shown in Figure 9(a) were obtained. From these loadings, the SES of individual cells was obtained. ｎ７１０ The percentages (scores) of the first and second principal components included in the SES are calculated, and the horizontal axis is the score of the first principal component, and the vertical axis is the score of the second principal component. The scatter plot shown in Figure 9(b) is obtained by plotting the score of the first principal component on the horizontal axis and the score of the second principal component on the vertical axis. For example, if the scores of both the first and second principal components are in the positive region (first quadrant of Figure 9(b)), ｎ７１０ can be expressed as the sum of the loadings of the first and second principal components, and the SES ｎ７１０ means that it can be expressed by subtracting the loading of the second principal component from the loading of the first principal component. ｎ７１０ means that the first and second principal components are hardly included. The black circles represent the SES of normal cells contained in normal specimens A to F. ｎ７１０The scores correspond to the SES scores of tumor cells in specimens G to I, which are suspected to be mesothelioma. ｎ７１０ The scores for normal cells are distributed mainly in the third and fourth quadrants, but also in the first and second quadrants, whereas the scores for tumor cells are concentrated in the first and second quadrants. This result indicates that normal cells and tumor cells, which cannot be clearly classified using the SES in Figures 5 and 7, can be numerically classified using the second principal component score.

[0044] Here, we will explain how to calculate the score of each principal component. The score of the first principal component of patient n (n = 1 to N) is shown in the following formula 1. As shown in the following formula 2, w １ (λ) is the weighting of the first principal component (eigenvector), and w １ Assuming that the sum of squares of (λ) is 1, the variance V of the score of the first principal component is １ becomes the maximum w １ (λ) is determined.

[0045]

[0046]

[0047]

[0048] The score of the second principal component for patient n (n = 1 to N) is shown in the following formula 4. As with the first principal component, w ２ (λ) is the weighting of the second principal component, as shown in the following equation 5, ２ The sum of squares of (λ) is set to 1. Furthermore, as shown in the following equation 6, ２ (λ) is w １ (λ) and the orthogonal condition is imposed, and under this condition the variance of the scores of the second principal component is maximized. ２ (λ) is determined.

[0049]

[0050]

[0051]

[0052] Then, for the score of the nth principal component of patient n (n = 1 to N), add the orthogonal conditions and find the score that maximizes the variance of the nth principal component.ｎ In this way, spectral data is much simpler than image data and allows for less arbitrary analysis.

[0053] For the scatter plot created as shown in Figure 9(b), the SES of unknown cells is used to determine whether they are normal cells or tumor cells. ｎ７１０ By comparing the SES obtained from an unknown cell, the type of the cell can be estimated. ｎ７１０ The loading scores of the first and second principal components shown in Figure 9(a) were calculated, and whether the cell was normal or tumorous was estimated based on where it was plotted in the diagram in Figure 9(b). For example, if the score of an unknown cell is in the third or fourth quadrant, where normal cells have a high score, the unknown cell is likely to be a normal cell, and if it is in the first or second quadrant, where tumor cells have a high score, the unknown cell is likely to be a tumor cell.

[0054] Methods for this two-class classification of normal cells or tumor cells include the k-nearest neighbor method, the decision tree method, and the random forest method. As shown in FIG. 10( a), the k-nearest neighbor method is a method of classifying cells by taking a majority vote among the k nearest samples. As shown in FIG. 10( b), the decision tree method is a method of creating partitions (decision trees) in a scatter plot and classifying cells by a majority vote within the partition. As shown in FIG. 10( c), the random forest method is a method of classifying cells by taking a majority vote among multiple different decision trees and then taking a majority vote of the majority vote results.

[0055] As shown in FIG. 11, the SES of normal cells contained in one of specimens A to F ｎ７１０ The test data is an unknown cell, and the SES of the remaining normal cells ｎ７１０ The data for predictive model training to create a scatter plot are the SES of tumor cells contained in one of the specimens G to I. ｎ７１０ The test data assumes that the unknown cells are SES of the remaining tumor cells. ｎ７１０The data were used as training data for creating a scatter plot, and the test data for normal cells and tumor cells were classified into two classes using the method described above. As a result, the accuracy rates of the k-nearest neighbor method, decision tree method, and random forest method were 69.5%, 70.9%, and 72.4%, respectively. ｎ７１０ This example demonstrates that machine learning can be used to distinguish between normal and tumor cells, even though the light scattering spectra of normal and tumor cells obtained from actual patients are extremely difficult to interpret from a physical perspective.

[0056] Although this analysis only takes into account the first and second principal components, for example, by considering higher-order principal components in the classification, it is possible to increase the accuracy rate of normal cell / tumor cell classification based on minute shape characteristics of the light scattering spectrum that cannot be visually detected by humans, thereby achieving higher accuracy. Furthermore, by analyzing the vector information of the dark-field image shown in Figure 3 and the frequency information obtained by two-dimensional Fourier expansion in the same manner as the light scattering spectrum and incorporating them into the learning model as features together with the light scattering spectrum, even higher accuracy can be expected.

[0057] Furthermore, in pathological diagnosis, the key issue is not determining whether individual cells contained in a pathological specimen are normal or abnormal, but diagnosing whether the pathological specimen contains abnormal cells. For example, even if the accuracy (accuracy rate) of individual cell discrimination is 70%, if discrimination is performed on several hundred cells contained in a cytological specimen and the results are statistically analyzed to diagnose the presence of abnormal cells in the pathological specimen with nearly 100% accuracy, the objective can be achieved at a high level. To improve the diagnostic accuracy of this pathological specimen, not only the discrimination accuracy of a single cell but also the measurement speed and measurement accuracy of the light scattering spectrum are important. Furthermore, if the overwhelmingly large number of normal cells can be removed in advance during measurement, and if there is estimated information on the number of abnormal cells contained in the specimen, it is believed that the diagnostic accuracy can be significantly improved. Thus, this example clearly demonstrates the possibility of achieving highly sensitive and highly accurate diagnosis of pathological specimens that surpasses conventional methods.

[0058] In this example, as in Example 2, pathological specimens were used in which normal cells and tumor cells obtained from actual patients were fixed and stained according to the existing pathological diagnosis procedure, and unlike Example 2, higher-order principal components were also taken into consideration, and the results of discriminating light scattering spectra by machine learning were shown. As in Example 2, normalized light scattering spectra (SES) for cells contained in specimens A to I were used. ｎ７１０ ) (see Figure 13) was subjected to principal component analysis, and the loadings of the first to tenth principal components were obtained. Figure 14 shows the contributions and loading results of the first principal component (PC1), the second principal component (PC2), ..., the tenth principal component. From these loadings, the SES of individual cells was obtained. ｎ７１０ The percentages (scores) of the nth principal component (n = 1 to 10) and the mth principal component (m = 1 to 10) included in the matrix were calculated, and a group of plots (scatter plots) with the score of the nth principal component on the horizontal axis and the score of the mth principal component on the vertical axis are shown in Figure 15. Figure 15 shows 100 scatter plots, including plots of the same principal component. Since swapping the vertical and horizontal axes of each principal component yields essentially the same information, 45 scatter plots (n < m) for the nth and mth principal components (top right of the matrix in Figure 15) were examined. It was found that the scatter plot of the scores of the second principal component (PC2) and the sixth principal component (PC6) can clearly distinguish between normal and tumor cells. While the scatter plot of the scores of the second and sixth principal components clearly shows the difference between normal and tumor cells, changing the cell type or adjusting the analysis parameters may result in clearer differences in other components.

[0059] As mentioned above, we found that the scatter plot of the scores of the second and sixth principal components can clearly distinguish between normal and tumor cells. On the other hand, we also suggest that the scatter plot of the scores of the second and sixth principal components may reflect the malignancy of tumor cells. That is, as shown in Figure 16 , the difference between tumor cells and normal cells is not simply a yes / no; intermediate states exist, and it is possible that intermediate states can be expressed as the malignancy of tumor cells. There are cells that were clearly diagnosed as mesothelioma by both the pathologist's assessment and the light scattering spectrum assessment (among these, cells with faded pigments were also identified), cells whose scores were near the origin and difficult to diagnose by the light scattering spectrum assessment, and cells that were clearly diagnosed as reactive mesothelioma by both the pathologist's assessment and the light scattering spectrum assessment. The malignancy of tumor cells can be expressed as increasing malignancy toward the upper left and decreasing malignancy toward the lower right.

[0060] By establishing training data in this way, when a pathologist is unsure of a diagnosis, they can refer to the discrimination results based on light scattering spectra to support pathological diagnosis using image data. Furthermore, if training data can be accumulated, the accuracy of discrimination using light scattering spectra can be further improved, ultimately leading to fully automated diagnosis without relying on pathologists. For example, as shown in Figure 17(1), light scattering spectrum data can be obtained from an unknown cell (pathologists make a diagnosis based on cell shape information of 1 μm or more using image data of the unknown cell), and information on intracellular structures of 1 μm or less can be obtained from the light scattering spectrum. Then, as shown in Figure 17(2), a score can be calculated from the light scattering spectrum data, compared with the training data, and the malignancy level of the cell can be calculated to support pathological diagnosis. Calculating the score using light scattering spectrum data is simpler than diagnosis using image data, and it enables highly sensitive and accurate discrimination of tumor cells. Furthermore, image data and light scattering spectrum data contain different information, which is expected to improve diagnostic accuracy in a complementary manner. Combining image training data with light scattering spectrum training data is expected to improve the accuracy of cell malignancy determination.

[0061] In this example, we show the results of distinguishing unstained living cells in a pathological specimen using light scattering spectra. Here, we show optical microscope images of two living cells, a Jurkat T cell line and a U937 monocyte cell line, and the normalized light scattering spectra (SES) of each cell. ｎ７１０ ) was obtained. It is extremely difficult to distinguish between these two types of cells based on their shape characteristics in the microscopic image shown in Figure 18(1). In the normalized light scattering spectra shown in Figure 18(2), there appears to be a difference in shape between the upper spectra, but this difference is not clear in the lower spectra.

[0062] The normalized light scattering spectrum (SES) shown in FIG. ｎ７１０ ) was subjected to principal component analysis to obtain the loadings of the first to tenth principal components, and the SES of individual cells was calculated. ｎ７１０ The scores for the nth principal component (n = 1 to 10) and the mth principal component (m = 1 to 10) were calculated, and the scatter plot shown in Figure 19 was obtained. Results showed that the scores for the two types of cells differed for almost all principal components, and the two types of cells were clearly separated for all principal components up to the 10th (PC10) principal component. This example produced better results than Examples 2 and 3, which used stained specimens, and clearly demonstrates that the present invention can also be applied to live cells. Furthermore, the clear distinction between unstained live cells substantiates the validity of the hypothesis presented in Example 1.

[0063] The present invention makes it possible to easily distinguish tumor cells, which have been difficult to distinguish by pathologists using optical microscope observations in the past, and may enable diagnosis that is not dependent on the experience or skill of the pathologist, and may also enable more accurate diagnosis. Furthermore, by systematizing the measurement of light scattering spectra, comprehensive characteristic information of cells can be obtained more sensitively and quickly than with optical microscope images, which may enable unmanned and faster pathological diagnosis.

[0064] REFERENCE SIGNS LIST 1 Microscopic scattering spectroscopy measurement system 2 Cell culture dish or specimen 3 Dark-field condenser lens 3' Bright-field condenser lens 4 Objective lens 5 Light source (halogen lamp) 6 Spectrometer 7 CMOS camera 10 Cell culture dish 11 Spacer 12 Culture medium 13 Cover glass 14 Cultured cells 20 Cell discrimination device 21 Scattering spectrum acquisition unit 22 Learning model 23 Unit for discriminating between normal and abnormal cells 31 Electromagnetic scattering spectrum measurement device 32 Scattering spectrum 33 Display device for discrimination results 41, 46 Excitation light 42 Sample (scatterer) 43 Mie scattered light and Rayleigh scattered light 44 Raman scattered light (or fluorescence) 45 Spectrometer 47 Optical filter (excitation light cut filter)

Claims

1. A cell discrimination method in which electromagnetic waves are irradiated onto a sample containing cells taken from a patient, the electromagnetic wave scattering spectrum from the sample is obtained, and a learning model that has been trained to learn the characteristics of the electromagnetic wave scattering spectra of normal and abnormal cells is used to distinguish between normal and abnormal cells.

2. A cell discrimination method in which a sample containing cells taken from a patient is irradiated with electromagnetic waves containing all wavelengths in a specific wavelength range, an electromagnetic wave scattering spectrum due to Rayleigh scattering and Mie scattering from the sample is obtained, and a learning model that has been trained on the electromagnetic wave scattering spectra of normal and abnormal cells is used to distinguish between normal and abnormal cells.

3. The cell discrimination method described in claim 1 or 2, wherein the electromagnetic wave scattering spectrum is two-dimensional data representing the relationship between the scattered electromagnetic wave intensity and the wavelength of the electromagnetic wave, and the learning model learns the two-dimensional data and classifies the differences between normal cells and abnormal cells.

4. The cell discrimination method described in claim 3, wherein the learning model extracts features of the two-dimensional data using principal component analysis, the principal components being extracted by diagonalizing the variance-covariance matrix of the scattering intensity at each wavelength, and classifying the differences in the features of normal cells and abnormal cells.

5. The cell discrimination method according to claim 4, wherein in the principal component analysis, the first principal component is an eigenvector with the largest eigenvalue and has the largest variance in scattering intensity; the second principal component is an eigenvector with the second largest eigenvalue, which is orthogonal to the first principal component and has the largest variance in scattering intensity; and the nth principal component is an eigenvector with the nth largest eigenvalue, which is orthogonal to the (n-1)th principal components and has the largest variance in scattering intensity.

6. The cell discrimination method according to claim 1 or 2, wherein the sample is fixed with alcohol or formalin and stained.

7. The cell discrimination method according to claim 1 or 2, wherein the sample is a living cell.

8. A cell discrimination method according to claim 1 or 2, wherein the electromagnetic wave scattering spectrum is a forward scattering spectrum.

9. The cell discrimination method according to claim 1 or 2, wherein the irradiated electromagnetic waves have a wavelength of 100 nm to 1 μm.

10. A cell discrimination method according to claim 1 or 2, wherein the irradiated electromagnetic wave is visible light.

11. The cell discrimination method according to claim 1, wherein the electromagnetic wave scattering spectrum is either or both of a scattering spectrum due to Rayleigh scattering and a scattering spectrum due to Mie scattering.

12. A cell discrimination method as described in claim 2 or 11, wherein the difference between the electromagnetic scattering spectra of normal cells and abnormal cells is due to an increase or decrease in scattering that occurs when changes in cell structure occur that are smaller or equal in size to the wavelength of visible light.

13. The method for distinguishing cells according to any one of claims 1, 2 and 11, wherein the abnormal cells are tumor cells.

14. The cell discrimination method according to claim 13, wherein the tumor cells are cancer cells.

15. The cell discrimination method according to claim 1 or 2, wherein the irradiated electromagnetic waves are ultraviolet rays or X-rays.

16. A cell discrimination method according to claim 1 or 2, wherein the irradiated electromagnetic waves are infrared rays.

17. A cell discrimination device comprising a scattering spectrum acquisition unit that irradiates electromagnetic waves onto a sample containing cells taken from a patient and acquires an electromagnetic wave scattering spectrum, a learning model that has been trained with the electromagnetic wave scattering spectra of normal cells and abnormal cells, and a discrimination unit that uses the learning model to discriminate between normal cells and abnormal cells.

18. A cell discrimination device comprising an irradiation unit that irradiates a sample containing cells taken from a patient with electromagnetic waves including all wavelengths in a specific wavelength range, an electromagnetic wave scattering spectrum acquisition unit that acquires Rayleigh scattering and Mie scattering, a learning model that has been trained with the electromagnetic wave scattering spectra of normal cells and abnormal cells, and a discrimination unit that uses the learning model to discriminate between normal cells and abnormal cells.

Images